Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
  • H04L1/0056—Systems characterized by the type of code used
  • H04L1/0057—Block codes

Definitions

• DIGITAL DATA DECODER THAT DERIVES CODEWORD ESTIMATES FROM SOFT DATA
• This invention relates to the field of forward error correction, and, more specifically, to an algebraic decoder capable of estimating codewords from soft data.
• Forward error correction is an abstract but vital field to today's communications. Much if not most of today's communications, including voice telecommunications, is transmitted as digital data. Forward error correction facilitates data communication by detecting and correcting data errors introduced during transmission. The underlying principle of all forward error correction is to add sufficient redundant data to detect and correct one or more errors over a predetermined amount of data. Forward error correction codes must therefore have sufficient redundant data to be useful, but not so much as to significantly slow the data transmission.
• encoders Before the data is transmitted, encoders add correction codes according to the selected encoding method, forming a "codeword" comprising a plurality of characters or symbols. For example, a codeword comprising eight binary characters (“symbols”) has eight "1 's” or “O's” to convey the information data and the correction code (redundant) data.
• Each character in the resulting codeword is modulated into a signal and transmitted.
• a demodulator receives the transmitted signal and decides what character the signal represents. Due to problems in transmission (fading and the like), a demodulator may not be able to make a definite decision whether an unclear signal represents one character or another.
• the demodulator makes a guess of the character and delivers the character (a "hard” symbol) to the decoder, as is known in the art.
• the demodulator produces a hard symbol, which belongs to the finite set of symbols that can be produced by the encoder.
• the demodulator produces a reliability (or "soft") value for that symbol, indicating the confidence level.
• Receiver performance is reflected in the percentage of received data blocks that are rejected by the receiver. In certain applications, where some delay is tolerable, rejected data blocks can be replaced by requesting retransmission by the transmitter. Time intensive applications must either attempt to use the soft data or attempt to work around discarded data blocks.
• the most time intensive application for data transmission is digitally encoded speech.
• speech applications the quality of the received signal is directly related to the accuracy of the data received over a given time period.
• Dropped data blocks cause signal distortion, missing syllables and gaps in the speech. Therefore, recovering soft data is important in time sensitive data transmissions, such as digitally encoded speech.
• One decoder that takes advantage of soft data is an "errors and erasures" decoder. This decoder repeatedly decodes a soft data block, where the least reliable characters are changed. The best answer among the ones obtained by the repeated decoding is selected according to appropriate criteria.
• the complexity of an errors and erasures technique increases with the number of iterations of decoding.
• the time required for the number of iterations and the complexity of the calculations necessary for an errors and erasures decoder generally precludes using this type of decoder in time-intensive data transmission, such as voice.
• An algebraic decoder uses a forward error correction code based on a generator matrix to encode and an additional parity check matrix to decode the information data.
• the demodulator produces a received word, which contains the hard data. Specifically, the received word has the same length as the transmitted codeword, and its symbols belong to the set of symbols produced by the encoder.
• the demodulator produces a vector of soft data, one value for each symbol in the received word.
• the decoder processes the received word and the soft values to produce a codeword estimate.
• the soft values act as a bias, making the decoder more likely to change symbols with low reliability in finding the codeword estimate, since such symbols are the ones most likely to be in error.
• the present invention distinguishes itself from an errors and erasures decoder in that it requires a single decoding pass. It is also capable of exploiting the value of the soft information more directly.
• the decoder operates over the field of the code, and does not use any extensive arithmetic operations over real or complex numbers. It can be implemented very efficiently in hardware, and requires very few operations when storage is available for pre-computed quantities.
• a decoder according to an exemplary embodiment of this invention accepts a received word comprising a plurality of characters. The decoder first produces a syndrome for the received word using the parity check matrix.
• the decoder produces a plurality of cosets by determining a syndrome (using the parity check matrix) for each member of the predetermined set of received words, and grouping all members of the set having the same syndrome into a coset.
• the decoder chooses a member of the coset corresponding to that syndrome as the most likely error pattern. That choice is biased by the reliability values.
• the chosen error pattern is then subtracted from the received word to produce the codeword estimate.
• the information symbols corresponding to the codeword estimate are then fed to the next stage, for example, a speech decoder.
• the biasing in the decoder is obtained by computing a modified Hamming weight incorporating the soft data.
• FIG. 1 is a block diagram of a mobile station in which this invention operates
• FIG. 2 is a block diagram of the receiver of FIG. 1 in which a method of this invention operates;
• FIG. 3 is a flow chart of processing according to one exemplary embodiment of the method of this invention.
• FIG. 4 is a flow chart of processing according to another exemplary embodiment of the method of this invention.
• FIG. 5 is a flow chart of processing according to a further exemplary embodiment of the method of this invention.
• FIG. 1 is a block diagram of a mobile station (also called a wireless telephone, cellular telephone or cell phone), shown generally at 10, that decodes soft information according to the method of this invention.
• Mobile station 10 includes an antenna 12 for sending and receiving radio signals between itself and a wireless network.
• Antenna 12 is connected to duplex filter 14, which enables receiver 16 and transmitter 18 to receive and broadcast (respectively) on the same antenna 12.
• Receiver 16 demodulates, demultiplexes and decodes the radio signals into one or more channels, as will be explained in more detail in connection with Fig. 2.
• Such channels include a control channel and a traffic channel for speech or data. Speech is delivered to speaker 20, data is delivered to a connector 21 to a modem or fax.
• Receiver 16 delivers messages from the control channel to processor 22.
• Processor 22 controls and coordinates the functioning of mobile station 10 responsive to messages on the control channel using programs and data stored in memory 24, so that mobile station 10 can operate within the wireless network.
• Processor 22 also controls the operation of mobile station 10 responsive to input from user interface 26.
• User interface 26 includes a keypad 28 as a user-input device and a display 30 to give the user information. Other devices are frequently included in user interface 26, such as lights and special purpose buttons.
• Processor 22 controls the operations of transmitter 18 and receiver 16 over control lines 34 and 36, respectively, responsive to control messages and user input.
• Microphone 32 receives speech signal input, converts the input into analog electrical signals and delivers the analog electrical signals to transmitter 18.
• Connector 21 receives digital data input from, for example, a fax machine or a modem.
• Transmitter 18 converts the analog electrical signals into digital data, encodes the data with error detection and correction information and multiplexes this data with control messages from processor 22. Transmitter 18 modulates this combined data stream and broadcasts the resultant radio signals to the wireless network through duplex filter 14 and antenna 12.
• Fig. 2 is a more detailed block diagram of the receiver 16 of Fig. 1.
• Demodulator 200 receives radio signals from duplex filter 14, and demodulates the radio signals into characters. At the same time, demodulator 200 determines a confidence score, as is known in the art, indicating how confident it is that each character is accurate.
• the decoded character is delivered to a code reassembler 202 according to an exemplary embodiment of this invention.
• Code reassembler 202 is advantageous in systems that use time slots and interleave encoded data as a hedge against fading in the transmission.
• the Global System for Mobile communication (GSM) standard for example, specifies such interleaving.
• Code reassembly 202 takes interleaved characters and reassembles them in their proper order.
• the output of code reassembly, a received word is delivered to a decoder 204 according to my invention.
• Decoder 204 receives a received word and a character reliability value for each character in the received word. Decoder 204 decodes the received word using precalculated syndrome-coset tables 206. According to an exemplary embodiment of this method, decoder 204 uses the character reliability value to select one of a plurality of syndrome-coset tables. Decoder 204 calculates the syndrome of the received word according to equation 2, below. The syndrome of the received word is used to select a coset leader from the selected coset leader table. The coset leader is subtracted from the codeword to derive a codeword estimate.
• Received words that are determined to be control messages are delivered to processor 22 (Fig. 1 ).
• Received words that are voice traffic are delivered to synthesizer 208, which uses the data to synthesize speech.
• the synthesized speech is delivered on analog wires to speaker 20. If mobile station 10 is being used as a data modem, then the information symbols are delivered directly from decoder to the data output 21.
• parity check matrix H is a full rank matrix selected so that
• T is the matrix transpose function.
• H is the matrix
• a received codeword z is transposed with the matrix H.
• s 0
• the codeword has no errors (per equation 1 ).
• s non-zero, an error has been detected.
• a Hamming weight is assigned to each received word z.
• the Hamming weight w(z) is defined as the number of non-zero characters in the received word z, wherein the Hamming weight increases with the number of non-zero characters. This weighting system is based on the fact that it is more likely that fewer characters were changed during transmission than more characters.
• a coset element with the smallest Hamming weight is chosen as the coset leader, e.
• the syndromes and the corresponding coset leaders are given in Table 1.
• the character reliability value ⁇ is a nonnegative integer that increases with increased character reliability.
• a new weight w' is computed for each coset by summing the character reliability values.
• a new coset leader e' is selected according to the new weight w'.
• a plurality of tables is developed for every possible combination of weights w'.
• the first four characters are more reliable than the last three. Therefore, the error correcting capability of the code is focused on the unreliable section of the received word (the last three characters).
• the syndromes and the corresponding new coset leaders are given in Table 2.
• Table 2 the rows have been rearranged by increasing value of w'. Note in particular that there are three new coset leaders. As compared to Table 1 , in Table 2 the 1 's in the coset leaders have shifted towards the right (towards the last three characters), where the less reliable characters are.
• Table 2
• Fig. 3 is a flow chart describing processing according to an exemplary method. This flow chart describes both the computations necessary to calculate a codeword estimate, and also to generate a table that can be referenced by the character reliability values r as may be used in the example of Fig. 2. Processing begins in box 300, when a received word and the reliability values for each character in the received word are accepted at decoder 204. Processing continues to box 302, where the syndrome of the received word is computed using the parity check matrix. In box 304, a syndrome is computed for each member of the set of possible received words.
• each member of the set of possible received words is classified into a coset based on its respective syndrome.
• the modified weight w' incorporating the reliability values is computed according to equation 2 for each of the cf elements of the coset identified by the syndrome obtained in box 302.
• the element e' with the smallest modified weight w'(e') is selected as the most likely error pattern.
• the codeword estimate is delivered in box 316.
• Fig. 4 is a flow chart of processing in decoder 204 when the coset leader- syndrome tables are precalculated and stored in coset tables 206.
• the coset leader-syndrome tables are calculated offline, according to steps 304 and 308, and the weights are calculated as described in the flowchart of Fig. 3. These tables are stored in memory (ROM, for example). Processing begins in box 400 when a received word and reliability values for each character in the received word are accepted. Processing continues to box 402, where a coset leader-syndrome table is selected based on the reliability values. In box 404, the syndrome for the received code word is calculated, and, in box 408, the coset leader with the same syndrome as the received word is subtracted from the received word. The derived codeword estimate is delivered in box 410.
• Fig. 5 is a further flow chart of processing in decoder 204 wherein the coset leader-syndrome tables are precalculated and stored, as in FIG. 4.
• Processing begins in box 500 where a received word and reliability values for each character in the received word are received.
• the syndrome for the received code word is calculated.
• processing continues to box 504, wherein the coset with the same syndrome as the received word is selected.
• the coset leader for that coset is selected based on the reliability values received with the received word.
• the coset leader with the same syndrome as the received word is subtracted from the received word.
• the derived codeword estimate is delivered in box 510. In this manner, decoder 204 uses its processing capacity more efficiently that in FIGS. 3 and 4, by performing the fewest calculations per received word.
• Bounded distance decoding is a variation on the above syndrome based decoder that uses a subset of the cf available coset leaders for decoding. For a given L ⁇ q M , the subset includes coset leaders of lowest weight. If the received word z produces a syndrome whose corresponding coset leader does not belong to the subset, then a decoding failure is declared. Different systems process failures differently. For instance, a retransmission is requested, or a block of information is erased.
• a bounded distance decoder is defined so that the L coset leaders in the subset are chosen according to the new weight w'.
• w' the effect of w' is not to produce new coset leaders, but only to modify their order, then a bounded distance decoder may still be different for w' than for w.
• a parity check matrix corresponding to G is given by:
• the syndromes, the coset leaders, and the coset leader weights are given in Table 5.
• the soft information ⁇ is assumed to be available for each received character z,.
• w' the new weight
• w' the new weight
• the syndromes and the new coset leaders are given in Table 6.
• the coset leaders have shifted their non-zero characters towards the left. Again, this means that the error correcting capability of the code is being focused on the unreliable section of the received word.
• coset leaders are ordered by increasing Hamming weight.
• a "radius" p is picked such that 2 p + 1 ⁇ d.
• L all coset leaders of Hamming weight ⁇ p are kept, and all other coset leaders are removed.
• the coset leaders are perfectly symmetric in that a permutation of a coset leader is another coset leader.
• d 3

Landscapes

• Engineering & Computer Science (AREA)
• Computer Networks & Wireless Communication (AREA)
• Signal Processing (AREA)
• Error Detection And Correction (AREA)
• Detection And Prevention Of Errors In Transmission (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=22289117

Family Applications (1)

Country Status (8)

Families Citing this family (149)

Family Cites Families (7)

• 1998
  • 1998-06-22 US US09/102,291 patent/US6145110A/en not_active Expired - Lifetime
• 1999
  • 1999-04-28 AU AU37695/99A patent/AU3769599A/en not_active Abandoned
  • 1999-04-28 WO PCT/US1999/009239 patent/WO1999067913A1/en not_active Application Discontinuation
  • 1999-04-28 EP EP99920123A patent/EP1090473A1/en not_active Withdrawn
  • 1999-04-28 CN CN99807773A patent/CN1126312C/zh not_active Expired - Fee Related
  • 1999-04-28 EE EEP200000762A patent/EE200000762A/xx unknown
  • 1999-04-28 BR BR9911427-5A patent/BR9911427A/pt not_active IP Right Cessation
• 2002
  • 2002-01-29 HK HK02100708A patent/HK1039007A1/xx not_active IP Right Cessation

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events